Yeshua taught in Aramaic. The Gospels reached us in Greek. This rendering starts from the Greek and asks, of each word, what Aramaic most likely stood beneath it and how surely that can be known. This page shows how: the sources we read, the engine that makes them answer that one question about every word, the ladder that grades each answer, and the checks a chapter passes before it is published.
Seven witnesses in, six surfaces out, with five essays beside them. Each claim in between carries its source and its tier. The stages below walk the same chart in detail.
None of them is treated as the original, and none is quietly promoted. Each enters under its own name, with its own limits on what it may claim.
The SBL Greek New Testament, a modern critical edition of the Greek. Every reading starts here, and every claim is keyed to a Greek word. The Greek is the base.
The Syriac Gospels, consulted for every word of all four Gospels. It is a witness to the Aramaic, not the original, and on its own it cannot settle a reading. Where the Galilean lexicons attest the same word, the Peshitta’s part is the shape of the sentence: the order of the words and their tenses. Where only the Peshitta has the word, it is shown as a Syriac witness and labelled so.
Two earlier Syriac witnesses. Burkitt’s 1904 text supplies a provisional reading under 831 verses, and the Sinaitic palimpsest in Lewis’s 1910 edition, re-read from the page and aligned by verse, is shown under 731 verses as an independent witness, placed to within about a verse. Both are shown beside the reading. The Burkitt text is not scored; the Sinaiticus witness enters the consensus as an independent, agreement-only witness that corroborates a word and does not move a tier by itself.
When Yeshua quotes the Torah or the prophets, the words underneath are Hebrew. The 89 Old Testament quotations render from the Hebrew source, with any divergence from the Greek disclosed rather than corrected.
Sokoloff’s Dictionary of Jewish Palestinian Aramaic, all 5,980 entries, joined to the corpus in two ways that are kept apart: by exact headword, where the dictionary has the same word, and by root family, where only the root is shared. Only an exact join receives the DJPA badge. By occurrence, 61.6% of the glossary’s words join an exact DJPA headword; ambiguous joins were screened by two models and settled by hand. The lexicon labels and corroborates a word; it does not choose the English or set the tier. Where a Galilean form is shown with its vowels, the vowels are anchored to a printed vocalization — Sokoloff’s entries or Dalman’s 1905 Grammatik, indexed page by page from the print — and the display says whether they are printed or reconstructed by regular morphology.
The Comprehensive Aramaic Lexicon, queried for all 1,845 distinct Aramaic words in the four Gospels, particles included. CAL files an entry for 1,756 of them. Where we can say which homograph is ours (604 of the 1,305 glossary words), the card carries the vocalized form, the part of speech, the root, and the numbered senses with the dialects that attest each one; where the consonants could be several unrelated words, the card says nothing. Attested in a dialect means the word is recorded in that dialect’s surviving texts, not that Yeshua spoke it. A high-confidence Common Aramaic match counts in the same Galilean and Western lexical-attestation lane because the family-wide word includes Galilean; the exception screen removes only a demonstrated same-sense case where Western Aramaic prefers another lemma, and no pronunciation is claimed. With explicit dialect tags and supporting Western citations, 380 of the 604 resolved words carry merged Western lexical attestation, including 229 in the Galilean lane. CPA is recorded but not counted as Western evidence: CAL’s Christian Palestinian texts are themselves a translation of the Gospels, so a CPA hit tells you the CPA translators used the word, not that a Galilean did. Used by permission of CAL’s maintainer and credited wherever it shows.
Published Aramaic reconstructions (Casey, Dalman, Burney, Torrey, Kutscher, Fitzmyer among them), each weighed on its evidence, and a minority position weighed as one. No single scholar is treated as the authority.
The core vocabulary, worked out one word at a time and written down so that later verses stay answerable to it: 35 governed words in the registry, and 37 more renderings tuned by hand. Malkuta is built on the verb to reign and names a reigning in progress, not a territory, so it renders reign. Tub is the ordinary verb to turn, or turn back, which is what the prophets ask for; repent is a later Latin overlay carrying penance the Aramaic does not say. The record keeps the reading tied to the root, verse after verse, instead of drifting toward whatever the familiar English happens to be.
One caution layer sits over the Syriac evidence: 21 documented patterns of Syriac translator habit (Williams, 2004) filter what a Syriac-only reading is allowed to claim. A Louw–Nida semantic-domain code sits on the Greek (1,296 words), a label for the renderings that read straight from the Greek.
The Lewis–Gibson 1899 Christian Palestinian Aramaic lectionary, the closest surviving relative of Galilean, is being transcribed page by page under a two-pass gate (183 of 318 pages as of 2 September 2026, the drive running to completion). Its lines are placed by verse with an aligner anchored on the Greek and Peshitta word order, and shown under 626 verses as a display-only Western witness, labelled by confidence.
Two further Palestinian lexical anchors sit beside the Galilean lexicon: Schulthess’s Lexicon Syropalaestinum (1903; 3,604 Christian Palestinian Aramaic entries, indexed from the print) and Dalman’s 1905 grammar. Both are reference lanes for forms and vocalization, not scoring witnesses. A second CPA Gospel witness, Lewis’s 1909 Codex Climaci Rescriptus, has been read under the same two-pass gate (22 of its 24 Gospel pages recorded as of 2 September 2026, shown as a display witness of 212 Gospel verses), and Phillips’s 2025 edition of CCR II and XI (Acts and epistles, not Gospels) after it; whether non-Gospel CPA may count as Western attestation is an open decision, so those lanes display and do not score.
Every source named here is cited and linked, edition by edition, on the sources page.
For each word we ask the same thing: what Aramaic most likely stood beneath this Greek word, and how surely do we know it? We record the same evidence for each word so that anyone can check the answer.
All four Gospels (89 chapters, 3,779 verses), word by word. Each word becomes a record of about 64 facts: the Greek, the Syriac, the Palestinian Aramaic where attested, the gloss, the tier, where each claim came from, and whether the word sits in speech or in the narrative frame.
A glossary entry exists only where an Aramaic root can be established, and it holds only the occurrences that share that root.
Each felt-sense entry is drafted ten times independently and merged. Where the drafts disagree, the disagreement stays visible.
The machine builds; a human reads. A small set of renderings, 37 words so far, is fine-tuned by hand for spoken fluidity and clarity of sense. Each is recorded with its reason and kept through every rebuild.
Each claim carries its source. Exact DJPA joins and accepted CAL Galilean or Common Aramaic matches share one Galilean/Western lexical-attestation class; no second Common Aramaic badge class is opened. Otherwise the card names a Peshitta or Syriac witness.
Where a witness is silent, the cell shows a · gap rather than a guess.
The evidence places each word on a four-rung ladder, and the rung decides which language governs the English. Where the evidence is strong, the Aramaic leads. Where it is thin, the Greek holds.
The fourth rung is about where a word sits, not how much evidence it has. Words inside reported speech do not fall to the fourth rung for lack of evidence; when the Aramaic beneath a spoken word cannot be pinned, it is T3 and says so. 12,520 words stand on the fourth rung, 10,141 of them with a Peshitta word shown beside them.
Repeated rounds of read-aloud revision across all 89 chapters, each decision logged with its reasons and with what it superseded.
214 of those decisions are encoded as machine-checked rules and enforced corpus-wide on every rebuild, so a rendering that breaks a settled decision is caught before it ships.
Flagged glossary senses are cross-examined by two independent AI reviewers. A fix lands only where both agree; where they disagree, the question goes to the human.
Each chapter is checked verse for verse against canonical versification: nothing missing, nothing repeated, nothing out of order, and the critical text’s known omissions are disclosed rather than papered over. The built editions are audited line by line, and the Audio script must be pure voice text: no verse numbers, no headings, only the words meant for the ear.
The reading text and its apparatus are built from the same data, cross-linked in both directions, and published together. What you read and what it rests on stay side by side.
Matthew, Mark, Luke, and John, rendered whole and confidence-ranked: read it, and press play to listen along, verse by verse.
Talk-along rows under every verse: Aramaic first, each evidence layer a toggle, each word carrying its tier. A living draft, and marked as one.
The felt-sense reference, grouped by Aramaic root, every occurrence linked back to the verse it lives in.
All four Gospels drawn as a single network of embodied Aramaic sense-domains.
English, Aramaic transliteration, Syriac, or Greek. Every hit linked to its interlinear verse.
Both editions of each Gospel as .docx, free to download and carry with you.
Why Aramaic at all · the four words beneath “save” · the Hebrew inside the quotations · how much Galilean is really here · whether he contradicts himself.
The surfaces are checked against each other on every build: each glossary occurrence links to its verse and each interlinear word to its glossary card, with no broken anchors allowed.
Everything above describes how the reading is built. This is where it stops.
✦ A way to hear the Gospels with Galilean ears
✦ A rendering built from the Greek text of the written Gospels, listening for the Aramaic beneath it
✦ Graded word by word on one four-tier ladder
✦ Open, and revised in the open
· The entire Bible: this is the four Gospels alone
· A claim to possess all of Yeshua’s exact words
· A claim that the Greek Gospels are wrong
· A secret original Gospel
· The Peshitta dressed up as the original
· Root-mysticism; nothing renders without a lexical anchor
· Finished. These are drafts, and they say so